Robot Vacuum Cleaner Personality and Behavior

Home , Robotic vacuum cleaner, Vacuum cleaner

Int J Soc Robot (2011) 3: 187–195 DOI 10.1007/s12369-010-0084-5

Bram Hendriks · Bernt Meerbeek · Stella Boess · Steffen Pauws · Marieke Sonneveld

Accepted: 14 November 2010 / Published online: 23 December 2010 © The Author(s) 2010. This article is published with open access at Springerlink.com

Abstract In this paper we report our study on the user ex- 1 Introduction perience of robot vacuum cleaner behavior. How do people want to experience this new type of cleaning appliance? In- 1.1 A New Kind of Presence terviews were conducted to elicit a desired robot vacuum cleaner personality. With this knowledge in mind, behavior Robot vacuum cleaners have been on the market for a few was designed for a future robot vacuum cleaner. A video years now and, to an increasing extent, these products are prototype was used to evaluate how people experienced the taking over our chores. Robot vacuum cleaners are vacuum behavior of this robot vacuum cleaner. The results indicate cleaners that clean ﬂoorings autonomously, and are among that people recognized the intended personality in the robot the ﬁrst service robots that enter our homes. This type of vacuum cleaner is a new kind of presence to get used to. behavior. We recommend using a personality model as a tool It exhibits autonomous behavior and it moves around our for developing robot behavior. homes and affects our daily lives. This is very different from conventional, non-robotic vacuum cleaners. People are no Keywords Robot vacuum cleaner · User experience · longer in full control of what the product is doing, where and Personality · Behavior when. People need to understand and trust the robot vacuum cleaner. Crucial for the user experience and acceptance of the product is the design of appropriate interaction between human and robot [1]. Although robot vacuum cleaners are a relatively new phenomenon, the idea in itself—having a robot cleaning assistant at home—is not. People have dreamt about it for decades. Back in the seventies Quasar Industries claimed to B. Hendriks () · S. Boess · M. Sonneveld have developed a robot capable of carrying out chores in Delft University of Technology, Delft, The Netherlands the home [2]. It turned out to be a fake. But today, more e-mail: [email protected] than thirty years later, a variety of robot vacuum cleaners is S. Boess on the market. These are developed by Electrolux, iRobot, e-mail: [email protected] Hanool, LG and Samung, among others [3]. Since robots M. Sonneveld are entering our homes now, it is interesting to study how e-mail: [email protected] users experience this new kind of presence in their domestic B. Hendriks · B. Meerbeek · S. Pauws environment. Philips Research, Eindhoven, The Netherlands B. Meerbeek 1.2 Personality e-mail: [email protected] S. Pauws People tend to behave towards artifacts in a social way, par- e-mail: [email protected] ticularly if artifacts exhibit some degree of autonomy such 188 Int J Soc Robot (2011) 3: 187–195

Fig. 1 From desired personality to evaluation of personality

as robot vacuum cleaners [4–9]. Aspects of anthropomor- Bartneck [1], and is a high-level description of the way phism—the attribution of human qualities to non-humans— in which the robot interacts with its environment. A well- are reported in studies on the experience of robot vacuum defined and clearly communicated personality can serve as cleaners. The most studied robot vacuum cleaner is the iRo- a mental model of the robot for users and so facilitate the bot Roomba. It is reported that people name it, and that peo- interaction [17]. ple ascribe a gender and personality to Roomba [4–9]. The Second, we investigate how people experience the per- personality of Roomba is sometimes described as stubborn, sonality of the robot vacuum cleaner as presented in the silly [5], crazy [6], intelligent [8] and dumb [9]. video prototype. Does the personality of the robot vacuum An accepted and widely used model that describes per- cleaner they experience match with the desired personal- sonality is the Five-Factor Model (FFM) of personality, ity? The relationship between the two parts of the study also referred to as the Big Five. This model organizes per- isshowninFig.1. A similar approach is suggested by sonality traits in terms of five basic dimensions: neuroti- Young, Hawkins, Sharlin and Igarashi: designers could use cism, extraversion, openness to experience, agreeableness robot expressions as a means to influence perception, in an and conscientiousness [10]. A number of instruments—or attempt to convey certain robot characteristics [7]. inventories—capable of measuring personality exist, among which: NEO PI-R [11] and TIPI [12]. Although these are inventories used to study personality in a systematic way, 2 Desired Robot Vacuum Cleaner Personality people assess personality on a daily basis and form an impression of others, in order to know what to expect [13]. 2.1 Research Method But in their daily lives, people do not only use human personality characteristics to describe people, they also use these to describe products [13] and interaction with products In order to find out what robot vacuum cleaner personal- [14]. Or more specific: interaction with computers [15], ro- ity people desire, a semi-structured interview was done with bots [16] and—as we have seen—even robot vacuum clean- six participants, two women and four men. They were se- ers [4–7]. Here, we refer to personality as the use of hu- lected because they were likely to be early adopters of ro- man personality characteristics to describe a robot vacuum bot vacuum cleaners. The participants share some distinc- cleaner. tive characteristics. They are all busy scheduled, and ei- Studies have shown that people prefer more extraverted ther have a background in technology or have affinity with and agreeable personalities over more introverted and for- technology. Besides this, they are Dutch. Before the actual mal ones when interacting with a computer [15] and a ro- interview, all participants were familiarized with the con- bot [16]. But these studies by Reeves and Nass [15], and cept of a robot vacuum cleaner by explaining the product Meerbeek, Hoonhout, Bingley and Terken [16] focus on ap- and presenting a visual overview of various robot vacuum plications with a social function, whereas a robotic vacuum cleaners. For the interview, thirty personality characteristics cleaner primarily serves service purposes. were adopted from existing Big Five personality inventories [11, 12]. For each of the five dimensions of the Big Five, six 1.3 Study Design and Research Questions characteristics that were considered relevant and interesting with respect to robot vacuuming cleaning were selected. Of This study consists of two parts. First, we investigate what these six characteristics, three have a positive connotation kind of robot vacuum cleaner personality people desire. towards the dimension, whereas the other three have a neg- How do people want the personality of their robot vacuum ative connotation. The characteristics were randomly pre- cleaner to be? This personality is then used to develop a sented on cards, and the participants were asked to evaluate video prototype—a mock-up in the form of film—of ro- the desirability. The personality inventories were used as a bot vacuum cleaner behavior. The personality serves as a starting point, and the selected characteristics were cues to design guideline, as suggested by Meerbeek, Saerbeck and talk about the desired robot personality and behavior. It was Int J Soc Robot (2011) 3: 187–195 189

Fig. 2 Extent to which the personality characteristics are desired

found that people could quickly and easily assess the pre- 2.2 Results sented characteristics, and that they could make imaginative descriptions of their desired robot vacuum cleaner’s person- We discuss the most relevant results with respect to the de- ality. An overview of the selected characteristics is presented sired personality here. The participants desired a calm ro- in Fig. 2. bot vacuum cleaner (6 participants out of 6); it should ex- 190 Int J Soc Robot (2011) 3: 187–195

Fig. 2 (Continued)

press that it is in control of the situation. Furthermore, they 3 Design of the Video Prototype wanted the robot vacuum cleaner to be cooperative in na- ture (5 participants). All of them desired an efficient robot The results from the interviews were used as a basis for vacuum cleaner (6 participants). They wanted it to like rou- the robot vacuum cleaner behavior. This behavior was pre- tines (5 participants), as vacuuming is very much a routine sented by means of a video prototype, which was recorded at job. The participants desired a polite robot vacuum cleaner the Philips Research ExperienceLab. A part of this observa- (5 participants), as well as one that behaves in a systematic tional laboratory is arranged as a living room [18], and was way (6). To summarize the above, the interviews made clear therefore used to mimic real human-robot interactions in a that participants did not desire fancy gadgetry, but a robot domestic setting. Paper-based methods are perfectly suited vacuum cleaner that fulfills its cleaning task. Figure 2 pro- to explore a first series of ideas for personality and behav- vides an overview of the results of the interview. The inter- ior. But these ideas are hard to capture on paper, especially views resulted in more than just quantitative data. It resulted the subtleties that are so important in interaction. Full im- in rich information on what a certain personality characteristic could mean for the robot behavior, according to the par- plementation of a working robot is often too costly and time ticipants. For example, a calm robot vacuum cleaner was consuming. Video prototyping on the other hand, is a very considered not to make a lot of noise and not to disturb un- suitable way for studying human-robot interaction [19] and necessarily. could lead to results that are comparable to those that could In addition to the characteristics as presented in Fig. 2, be obtained from live interactions with the robot [20]. In a people also brought up additional desired and undesired ro- study by Walters, Syrdal, Dautenhahn, Boekhorst and Koay bot vacuum cleaner personality characteristics. Examples of [20] the results obtained from participants watching another these additional characteristics are grouped by dimension person interacting with a robot were statistically compara- and listed in Table 1. ble to those obtained from participants who interacted with Int J Soc Robot (2011) 3: 187–195 191

Table 1 Personality dimensions along with examples Dimension Characteristic Dimension Characteristic of characteristics brought up by participants Neuroticism Not depressed Agreeableness Helpful Timid Helpfullness Extraversion Cool Kind Not cuddlesome Not irritating Not too present Not sneaky Shy Self-willed Silent Serving Solitary Conscientiousness Hard-working Openness to experiences Dumb Neat No-nonsense

a real robot. With a video prototype it is impossible however, Table 2 Modalities and dimensions of expression, including examples to study long-term interaction experiences [4]. Modality Dimension Examples of effects The five-minute video prototype comprises a conceptual robot vacuum cleaner that encounters a variety of situations Motion Spatiality Directions, paths and planes, big–small at home. Situations such as recharging the batteries and vac- Time Fast–slow, accelerating–decelerating, uuming dirty spots. The many situations a robot vacuum regular–irregular cleaner may encounter at home trigger different behavior, Force Tension–release, control–uncontrolled depending on its personality. These possible situations were Sound Loudness Intensity, high–low discussed in focus groups—groups of possible end users of Pitch Frequency, high–low the robot vacuum cleaner—and a set of ten situations was Timbre Tonal–noise selected for implementation in the video prototype. In each Light Time On–off, fast–slow, regular–irregular of these situations the behavior of the robot vacuum cleaner is based on the desired personality characteristics. The translation from personality to behavior was inspired by a role play in which a group of actors was asked to act like The dimensions are based on studies by Klooster and Over- a robot vacuum cleaner with these desired characteristics. beeke [22], and Van Egmond [23]. Motion is defined by its Role playing was used as a way of accessing the experience dimensions of spatiality, time and force. Sound by loudness, of an interaction. Its use contributes to the quality of inter- pitch and timbre. Light is defined by time. Table 2 was used action in a product to be designed [21]. A similar process— from a desired robot vacuum cleaner personality to behavior as a way to describe behavior in a systematic way. The role through role playing—was followed by Meerbeek, Saerbeck play—which was captured on film—was analyzed, and for and Bartneck [1]. Attributes, such as macaroni, were avail- every situation the behavior was defined and described by able to support acting out some of the situations (e.g. ‘clean- means of the modalities and dimensions as listed in the ta- ing a dirty spot’). An introductory exercise was meant to ble. familiarize the actors with the personality. Then, the actors The behavior of the actors was adopted in an abstract were asked to act out situations—as if they were the robot way, as, of course, human behavior cannot be translated to a vacuum cleaner—making use of motion and sound (expres- robot vacuum cleaner directly. For example, one of the ac- sion through light was taken into consideration only after tors who was asked to act like a robot vacuuming a dirty this exercise). In general, the actors either crawled about or spot, started to clean that spot slowly thereby making repet- walked around at a slow pace to imitate a vacuum cleaner. itive, firm movements back and forth. These motion aspects Often, a typical vacuuming sound was simulated by them. of the behavior were then described as taking place on a In this study, the robot vacuum cleaner behavior focuses small plane, slow, regular and tensed. on expression through motion, sound and light. These ways These descriptions were then used for the development of expression seem most relevant for robot vacuum cleaner of variations of robot vacuum cleaner behavior. Videos of behavior. The different modalities all add up to the experi- variations of robot vacuum cleaner behavior were evaluated ence of the behavior, and strengthen each other. See Table 2, by six possible end users of the product. The participants for an overview of the modalities and dimensions of these were asked, while keeping the desired personality in mind, modalities, along with some examples of possible effects. which variations of behavior they preferred over another 192 Int J Soc Robot (2011) 3: 187–195

Fig. 3 Impression of the video prototype

and to indicate why they did so. Improvements for the de- to the robot as seen in the video. All participants completed signed behaviors were brought up by the participants too. this task individually, after having seen the complete video The preferred and improved behaviors were combined in a prototype. final video prototype of robot vacuum cleaner behavior on the basis of the desired robot personality. 4.2 Results The robot vacuum cleaner in the video is represented by When the fifteen participants were asked for their overall simply a cardboard box, as in this study the focus is on be- impression of the robot vacuum cleaner behavior, they de- havior. Elements, such as an on-off button, were added to scribed it by using personality characteristics such as appro- the appearance of the robot vacuum cleaner, as these were priate (3 participants out of 15), calm (3), boring (2), care- thought to be essential for the behavior as shown in the video ful (2), and systematic (2). It was probably experienced as prototype. An iRobot Roomba robot vacuum cleaner was boring, as the robot simply vacuumed the living room in a used as a test platform for the motion aspects of the behav- systematic way. Two participants worried about the feasi- ior to be designed. This Roomba was manually controlled bility of such a robot vacuum cleaner. Another said that if by means of a joystick via a Bluetooth connection. A micro- it could be trusted with respect to cleaning performance, it controller was used to control the light, which consists of six would be a great solution as it would really take over the separate LEDs. The vacuuming sound of the robot vacuum cleaning job. Signs of anthropomorphism and personality cleaner was recorded from a conventional vacuum cleaner, were observed throughout the evaluations. One participant whereas additional sounds were designed by using sound explicitly mentioned that he experienced the robot vacuum development software. See Fig. 3 for an impression of the cleaner as having a distinctive character. Out of fifteen par- video prototype. ticipants, fourteen assigned a gender to the robot vacuum cleaner. When talking about it, they frequently referred to it as ‘he’ or ‘him’. Only one participant consequently used 4 Experienced Robot Vacuum Cleaner Personality ‘the robot vacuum cleaner’ or ‘it’. Participants said that it felt as if the robot vacuum cleaner was alive (3 participants), 4.1 Research Method and like a domestic animal or a dog (3), or even like an in- fant (1). While evaluating the video prototype, we looked for signs The results with respect to the task of rating the person- of anthropomorphism and the perception of personality in ality characteristics indicate that the perceived personality particular. We therefore asked fifteen participants—eight matches with the intended product personality. Of the fif- women and seven men who fit into the same target group teen participants, twelve perceived the robot vacuum cleaner as described above—to think out loud while they watched as calm. They agreed with perceiving this characteristic the video prototype, and to describe their overall impression (10 participants), or even strongly agreed (2). Cooperative of the robot vacuum cleaner behavior afterwards. To inves- yielded similar results (10 participants agreed, 2 strongly tigate whether the perceived personality matched with the agreed). Again, twelve participants thought that the vac- intended personality, we asked the participants to indicate uum cleaner behaved in a systematic way (4 participants what kind of personality the robot in the video prototype ex- agreed, whereas 8 strongly agreed). And another twelve hibits. For each of the thirty personality characteristics (the thought that it liked routines (4 participants agreed, whereas same as those used during the interviews), participants indi- 8 strongly agreed). Then, the robot vacuum cleaner was gen- cated on a rating scale, ranging from 1 (strongly disagree) to erally perceived as being polite (7 participants agreed, 2 oth- 5 (strongly agree), to what extent they thought these applied ers strongly agreed). The participants were less outspoken Int J Soc Robot (2011) 3: 187–195 193

Fig. 4 The results of six of the characteristics

with respect to how efﬁcient they perceived the robot vac- as intended by the designer. These effects deal with the fact uum cleaner. Figure 4 provides an overview of the results that the video prototype is limited in time, and limited in sit- for these six personality characteristics. uations. Not everything a robot vacuum cleaner might en- Some of the responses by the participants were due to counter could be implemented. But some of the remarks prototype effects. These are deviations from the original idea went beyond the limited set of situations. What if there is 194 Int J Soc Robot (2011) 3: 187–195 nobody to help to empty the robot? What if it sucked up doing a longitudinal study on the user experience with a real jewelry? And those situations that were selected for imple- robot vacuum cleaner, in a real domestic setting. mentation happen in rapid succession; a complete vacuum cleaning cycle in only ﬁve minutes is rather quick for a real Open Access This article is distributed under the terms of the Cre- robot vacuum cleaner. This led some of participants to com- ative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, ment upon the robot vacuum cleaner’s ‘amazing’ recharging provided the original author(s) and source are credited. capabilities for example.

References 5 Conclusion and Discussion 1. Meerbeek B, Saerbeck M, Bartneck C (2009) Iterative de- People anthropomorphize robot vacuum cleaners and, as an sign process for robots with personality. In: Proceedings of aspect of this, they attribute personality characteristics. De- the new frontiers in human-robot interaction, symposium at the signers can make use of this phenomenon by deliberately AISB2009 convention, SSAISB. Springer, Berlin, pp 94–101. ISBN-190295680X designing a robot vacuum cleaner personality. This, by go- 2. Rutland J (1978) Exploring the world of robots. Pan Books, Lon- ing through a process of determining personality, developing don, pp 20–21 behavior in an iterative way and evaluating the user experi- 3. Prassler E, Kosuge K (2008) Domestic robots. In: Siciliano B, ence of the robot vacuum cleaner. We recommend using a Khatib O (eds) Springer handbook of robotics. Springer, Berlin, pp 1253–1281 personality model as a tool for developing robot behavior. 4. Sung J, Grinter RE, Christensen HI (2010) Domestic robot ecol- We found it useful as a guideline when taking design deci- ogy: an initial framework to unpack long-term acceptance of ro- sions and helpful in developing consistent behavior. bots at home. Int J Soc Robot 2:417–429 We have investigated what kind of personality is desired 5. Sung J, Grinter RE, Christensen HI, Guo L (2008) Housewives or technophiles?: understanding domestic robot owners. In: Proceed- for a robotic vacuum cleaner and found that people prefer ings of the 3rd ACM/IEEE international conference on human- a calm, polite, and cooperative robot vacuum cleaner that robot interaction 2008, Amsterdam, The Netherlands, pp 129–136 works efficiently, systematically and likes routines. Based 6. Sung J, Guo L, Grinter RE, Christensen HI (2007) “My Roomba is on these findings, we developed its behavior and evaluated it Rambo”: intimate home appliances. J Ubi Comput 4717:145–162 7. Young JE, Hawkins R, Sharlin E, Igarashi T (2009) Toward ac- with end users by means of a video prototype. The results of ceptable domestic robots: applying insights from social psychol- our evaluation indicate that people recognized the intended ogy. Int J Soc Robot 1:95–108 personality in the robot’s behavior. 8. Forlizzi J (2007) How robotic products become social products: We think that the designed personality is beneficial with an ethnographic study of cleaning in the home. In: Proceedings of the 2nd ACM/IEEE international conference on human-robot respect to the user experience, as it helps people to form a interaction 2007, pp 129–136 conceptual model of the robot vacuum cleaner. It leads to 9. Forlizzi J, DiSalvo C (2006) Service robots in the domestic en- a consistent set of behaviors, across situations over time. vironment: a study of the Roomba vacuum in the home. In: Pro- When users learn to know the personality of their robot, ceedings of the ACM/IEEE conference on human-robot interaction 2006, Salt Lake City, Utah, USA, pp 265–285 they are able to interact in an appropriate way and to pre- 10. McCrae RR, Costa PT (1987) Validation of the five-factor model dict how the robot responds. Furthermore, it is important to across instruments and observers. J Pers Soc Psychol 52:81–90 determine a desired robot vacuum cleaner personality, and 11. Costa PT, McCrae RR (1992) NEO PI-R professional manual. implement—from a user point of view—desired behavior. Psychological Assessment Resources, Odessa 12. Gosling SD, Rentfrow PJ, Swann WB (2003) A very brief measure For the user experience it is important to focus on the sub- of the Big Five personality domains. J Res Pers 37:504–528 jective aspects of human-robot interaction (such as the per- 13. Govers PCM (2004) Product personality. PhD, Delft University of ception of calm, cooperative behavior), instead of consider- Technology ing objective measures such as vacuuming speed and effi- 14. Órtiz-Nicolás JC (2006) Product personality in interaction. MSc, ciency only. This does not mean that cleaning performance Delft University of Technology 15. Reeves B, Nass C (1996) The media equation. Cambridge Univer- is unimportant. As we have seen, people expect a robot vac- sity Press, Cambridge uum cleaner to fulfill a user need: having a clean floor. 16. Meerbeek BW, Hoonhout HCM, Bingley P, Terken J (2008) The In future research, we want to investigate the effect of influence of robot personality on perceived and preferred level of a deliberately designed robot personality on the user experi- user control. Interact Stud 9(2):204–229 17. Norman DA (2004) Emotional design: why we hate (or love) ence in more detail. Do people have a good understanding of everyday things. Basic Books, New York the robot? Do people have trust in the robot? Also, we want 18. Van Loenen E, De Ruyter B, Teeven V (2006) ExperienceLab: to compare multiple personalities and the effect thereof on facilities. In: Aarts E, Diederiks E (eds) Ambient lifestyle: from the user experience of robot vacuum cleaner behavior. And, concept to experience. BIS, Amsterdam, pp 47–53 19. Dautenhahn K (2007) Methodology and themes of human-robot as watching a video prototype of robot vacuum cleaner be- interaction: a growing research field. Int J Adv Robot Syst havior is not the same as living with such a robot, we suggest 4(1):103–108 Int J Soc Robot (2011) 3: 187–195 195

20. Walters ML, Syrdal DS, Dautenhahn K, TeBoekhorst R, Koay KL Bernt Meerbeek holds a MSc in Information Management from (2008) Avoiding the uncanny valley: robot appearance, personality Tilburg University and a PDEng in User-System Interaction from Eind- and consistency of behavior in an attention-seeking home scenario hoven University of Technology. Since 2005, he works at Philips Re- for a robot companion. Auton Robots 24(2):159–178 search on user-system interaction in various projects—including iCat 21. Boess SU (2008) First steps in role playing. In: Proceedings of the and robot vacuum cleaners—and is highly involved in User Experience ACM conference on human factors in computing systems, Flo- research at the ExperienceLab facility. rence, Italy, pp 2017–2024 Stella Boess is assistant professor at the Delft University of Technol- 22. Klooster S, Overbeeke CJ (2005) Designing products as an inte- ogy. She teaches user research in and for design. She graduated in prod- gral part of choreography of interaction: the product’s form as an uct design at the HDK Saar and received her PhD degree at Stafford- integral part of movement. In: Proceedings of the conference on shire University. design and semantics of form and movement 2005, pp 23–35 23. Van Egmond R (2005) Emotional experience of frequency modu- Steffen Pauws holds a MSc in Medical Informatics from Leiden Uni- lated sounds: implications for the design of alarm sounds. In: de versity and a PDEng in Software Technology and a PhD, both from Waard D, Brookhuis KA, Weikert CM (eds) Human factors in de- Technical University of Eindhoven, the Netherlands. Since 2000, he sign. Shaker, Maastricht, pp 1–12 has been with Philips Research as a senior scientist covering various research areas including algorithms and user interfaces for music re- trieval. Bram Hendriks obtained his MSc Design for Interaction degree at the Faculty of Industrial Design Engineering of the Delft University of Marieke Sonneveld is assistant professor at the Delft University of Technology. He also studied Industrial Design at the Eindhoven Uni- Technology and has her own studio: Tactile Affairs. She graduated and versity of Technology and the Politecnico di Milano. For his graduation received her PhD at the Delft University of Technology. Her research project at Philips Research, he studied the user experience of a robot domain is the tactual experience in human product interaction, espe- vacuum cleaner. He has a strong user focus and is interested in com- cially the affective aspects of this experience in the context of health bining insights from design, engineering and the social sciences into and well-being. new solutions.